Digital computer systems are used in a number of applications in which virtually continuous availability of data is important to the operation of businesses or other entities using the systems. Generally, computer centers will periodically produce back-up copies of data on their various digital computer systems. Such back-up copies are usually not maintained on a continuous basis, but instead at particular points in time, often at night when the normal processing load on the computer centers is reduced and modifications to the data being backed up may be minimized, and in any case represent the data at the particular points in time at which the back-up copies are generated. Accordingly, if a failure occurs between back-ups, data which has been received and processed by the digital computer systems since the last back-up copy was produced, may be lost. Typically, such back-up copies will be maintained by the computer centers at their respective sites so that they may be used in the event of a failure, although some off-site archival back-ups may be maintained. Significant additional problems arise in the case of, for example, catastrophic events that can occur, such as may result from, for example, fire, flood or other natural disasters, intentional tampering or sabotage and the like, which may result in unintentional or intentional damage to an entire site or some significant portion thereof, since some or all of the back-up copies may also be damaged and the data contained thereon may be unavailable.
One problem that arises in connection with backing up information is that generally, during the backup operation, the mass storage subsystem from which information is being backed up is generally unavailable for the duration of the backup operation. This is generally necessary since typically it is desired to have the backup reflect the state of the information stored on the mass storage subsystem at a particular point in time, such as at the beginning of the backup operation, so that the backup will provide a "snapshot" of the information at that point in time. If the mass storage subsystem were available during a backup operation, the information would likely be updated during the backup operation, in which case it would be impossible to determine whether the backed up information represented the state of the information at the beginning of the backup operation, at the end, or sometime in between.
To alleviate this problem, mass storage subsystems have been developed which maintain a number of copies of the information in a mirrored arrangement. An illustrative such subsystem is described in U.S. patent application Ser. No. 08/820,912, filed Mar. 19, 1997 in the name of Philip Tamer, et al., entitled RDF-Based and MMF-Based Backups now U.S. Pat. No. 6,035,412, and assigned to the assignee of the present invention. In such a subsystem, at the beginning of a backup operation, one of the mirrored copies is separated from the others and is used as the copy to provide information for the backup. During the backup operation, the other copies (that is, the copies which are not used to provide information for the backup) may be updated, but the copy which provides information for the backup is not updated. Accordingly, the copy which provides the information for the backup will remain in the state in which it was at the beginning of the backup operation during the entire backup operation. At the end of the backup operation, the copy which provided information for the backup will be re-synchronized with the other copies, and during the backup operation, information as to the updates which were made to the other copies will be maintained for use in the re-synchronization.
One problem that arises in connection with such an arrangement is that it requires maintaining a number of mirrored copies to be maintained. Preferably, the mass storage subsystem will need to maintain at least two mirrored copies in an "online" state at all times, even during a backup operation. Accordingly, a third copy will also need to be maintained which can be used in the backup. A further problem can arise if it is desired to have the backup copies represent snapshots of the information stored on the mass storage subsystem at different points in time, which may require backup operations to overlap. This can occur, for example, if backup operations are initiated at desired time intervals, which are less than the amount of time required to complete each backup operation.